Title

Simulating The Navigation And Control Of Autonomous Agents

Keywords

Group Behaviour and Goals; Intelligent Agents; Machine Learning; Navigation; Robotics; Simulation

Abstract

While traditional data fusion started with systems which exploit the output of multiple sensors so as to optimise the characterisation or recognition of objects of interest, modem information fusion systems will increasingly integrate all types of information, including behavioural information and information resulting from modelling, analysis and computation. In many critical applications, modelling the behaviour of groups of coordinated autonomous entities must be carried out within physically accurate settings in order to provide realistic information about their likely behaviour. The simulated entities must conduct autonomous actions which are realistic, which follow plans of action, but which also exhibit intelligent reactive behaviour in response to unforeseen conditions. In this paper we describe how a complex and simulation environment can be used to fuse information about the behaviour of groups of objects of interest. The fused information includes the objects' individual pursuits and aims, the physical and geographic setting within which they act, and their collective social behaviour. The group control algorithms combine reinforcement learning, social potential fields and imitation. We summarise the design of a simulation system that we have designed based on these principles.

Publication Date

11-2-2004

Publication Title

Proceedings of the Seventh International Conference on Information Fusion, FUSION 2004

Volume

1

Number of Pages

183-189

Document Type

Article; Proceedings Paper

Personal Identifier

scopus

Socpus ID

6344294731 (Scopus)

Source API URL

https://api.elsevier.com/content/abstract/scopus_id/6344294731

This document is currently not available here.

Share

COinS